Front page | perl.beginners |
Postings from August 2012
On 2012-07-30 15:47, punit jain wrote:
> my $pm = new Parallel::ForkManager(10);
>
> my $count=0;
>
> foreach my $user (@users) {
>
> $pm->start($user) and next;
>
> my $result;
>
> --- do some processing ---
>
> $pm->finish(0, \$result);
>
> }
>
>
>
> $pm->wait_all_children;
>
>
> However the final value of count is not correct. Is there some race
> condition on same variable updation by the processes ?
>
> pm -> run_on_finish (
>
> sub {
>
> my $result = @;
>
> if (defined($result)) {
>
> my $count += $result;
>
> }
>
> }
I wonder why you expect a parent variable to be available to a child.
I am ignoring the bugs and style issues in your code, because they are
not relevant to that point.
If you need to merge results from child-processes, it is best to let
those child-processes write to unshared resources. Rows with an
autoincrement-id in a database are often fine for that. A separate file
per child-process on disk can also be fine.
After all children exited, the parent collects all those results and
merges them.
Map: the parent divides the tasks over the children
Reduce: each child reduces its assigned input to a result
Merge: the parent combines all the results into the final result
--
Ruud
Thread Previous
|
Thread Next